John Carmack's QuakeCon 2005 keynote @ The Tech Report

[Steve]

John Carmack speaks, gamers listen. It's that simple. So, what's he been saying at this year's QuakeCon?

Carmack started out by offering his assessment of the current state of PC gaming hardware, noting that he is largely satisfied with current trends. He marveled for a moment over the advances made in recent years, especially in graphics, and looked forward to continued progress on most fronts. Ever the PC guy, Carmack observed that the upcoming generation of game consoles looks to be very powerful, but said that PCs will, of course, be much more powerful than these consoles thanks to PC hardware's rate of progress.

[The Tech Report]

[Dougal]

I'm REALLY gonna save up me pennies for a 360 now.

[Stoo]

Heh, I liked the bit where he mentioned about graphics improving gameplay whereas physics etc won't..

Did he not play Doom3 then?

[dangel]

I think he has a point - how on earth do you scale physics?

[Stoo]

Well, he's kinda missing the point, with a PPU, if offloads physics from the cpu, letting it do other things like run better AI, have more elements on-screen, etc etc

Good old Carmack seems to have forgotten that having killer graphics isn't the be-all and end all in gaming, killer gameplay is..

After all, in Doom3.. yeah, graphical innovations there may have been, but they didn't help the gameplay at all, which IMO was really quite terrible.

[dangel]

Do we need a seperate card (200-300 quid) just for that? Given the level of physics in something like half life 2, isn't a dual core PC going to cope pretty well in the future?

And, again, how will you _scale_ it - with graphics cards the detail levels can be shifted, but with physics it's quite different and tends to be integral to the gameplay (can you push stuff around, knock stuff over to solve puzzles). I wouldn't disagree that gameplay is number one but I wonder how AI and physics can be appropriately scaled without forcing major alterations in gameplay. I think he also mentioned that physics isn't parallelized like graphics for processing (which is why graphics cards are so quick at what they do compared to a generalized CPU).

[Stoo]

he can argue as much as he likes, the facts are plain to see, a game using a PPU to offload that aspect from the CPU runs _much_ quicker than when it isn't used.

Of course you can scale physics, you can have more and more objects that react and use the physics engine, you can scale the precision of the physics, you can create soft-body dynamics, realistic environmental effects (Wind affecting bullet/rocket trajectory for example), etc. Of course, that's not all going to happen overnight with the first generation, but it's a step in the right direction.

The only limitation with physics acceleration is your imagination, something which Mr Carmack appears to be lacking these days.

[dangel]

I think you're being a bit harsh - and it's not as though mr carmack has too much impact on gameplay thesedays - he's all about engines (and, as an engine, doom 3 is impressive regardless of the shortcomings in content). I still (as a programmer) have a lot of respect for the guy - respect that's well due IMHO.

My point with scaling is that you _can't_ do it for elements that effect gameplay - no without producing a massively different game for one system to another. I suppose you could do so for say the bad guys being shot and have a lot more 'precision' in the way their bodies are catapaulted around by a rocket launcher but that's not going to sell more copies of a game. I think the best example we have of physics-driven gameplay atm is half life 2 and that's really opened our eyes to the possibilities. But would the experience been enhanced by ancillary objects (i.e. not prime factors in the path through the game) being more 'bouncy' (and now i'm thinking about beach volleyball, bad man that i am) or more reactive? Perhaps, but anything that's prime to the path through the game needs to work on alll systems and simply upping precision/speed on such objects may not be enough to make me part with hundreds of bucks on a PPU (as well as 2 GPUs, a cpu etc etc). I like the idea of wind - but here's a better example - wind blowing through the trees or grass would be pretty cool to see (but again makes no real impact on gameplay).

Sorry, i'm not attempting to be a downer on physics - given my profession that would be quite odd actually But i wonder with the advent of multi core systems whether we _need_ a PPU (or want to pay for it) and how such processing fits on in-order processing systems.

[Stoo]

I think it's in pretty much the same place when the initial 3D accelerators were announced, nobody could really see why they needed one, until the cards actually came out and the advantages were clear.

Anything that removes load from the CPU can only be a good thing, a multicore CPU may be able to give you the same performance as a uniprocessor + PPU, but a multicore CPU + PPU will be better than than a multicore processor alone, simply because the CPU isn't doing any of the calculations.

Sure they may not brilliant, but it's only the first iteration of the hardware, imagine what would have happened if we'd have given up on 3D acceleration after the first wave of cards?

Eventually I'd like to see PPU's being integrated into graphics cards, and then eventually into the cores, as I think that's where we'd see the most benefit.
Imagine the levels of realism that could be achieved through nth generation PPU's, perhaps a racing game where levels of grip are properly calculated as a combination of downforce, mechanical grip, ground effect, tyre compounds, road surface, environmental effects etc. Flight Sims could be taken to a whole new level, sports sims which actually had realistic environments, etc

I dunno, I'm not a games developer, but the prospect of a proper virtual world starts to become a real possibility, rather than the facades we currently play.

Real granularity, higher realism, hell even better blood splatters

The PPU looking after where an object appears in 3D space, all the object's attributes, and it's effects on other objects, and other objects on it etc etc

Yes, it would mean a shift in how games engines are thought about, but who wants to have the same thing re-hashed year after year? (Apart from EA..)

(This may have been a bit of a ramble)

[dangel]

(This may have been a bit of a ramble)

This whole thread is

Yeah i take your points on board and i think in time we'll see an answer to the questions raised. I guess, as you've said, it's going to take a while to bleed through into games - not just from a development prospective but also to get an installed base of the cards in people's machines. I just hope they aren't going to cost more than say, a good sound card!

Then again, the prospect of proper bouncy boobies has me hand on credit card

[Stoo]

Lowest common denominator

[dangel]

Lowest common denominator

Everyone likes boobies?

[Dougal]

Everyone likes boobies?

Not when they are too big to hold.

[Purple]

Properly modeled bouncy boobies would persuade me to buy a physics card. Seriously though, i would consider getting one of these cards in the future in order to SAVE money on an upgrade. Instaed of having to spend 500 quid on a full Mobo, RAM, CPU and monster GPU upgrade, i simple one off payment of 180 quid for a physics card could boost game performance enough to put off the massive upgrade for a few more months.